Search papers, labs, and topics across Lattice.
1
0
2
LRMs can be made both more accurate AND more efficient, without training, by dynamically steering their reasoning trajectories based on real-time confidence metrics.
